scvi-tools

Apply scvi-tools probabilistic models to single-cell omics data in AnnData format.

52|6|Updated Nov 24, 2025

Search Tags:#multimodal #single-cell #batch-correction #scRNA-seq #scvi-tools #SCANVI #scVI

Ready to run

npx skills add https://github.com/ovachiever/droid-tings --skill scvi-tools-ovachiever

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: scvi-tools
Source: https://github.com/ovachiever/droid-tings/tree/main/skills/scvi-tools
Command: npx skills add https://github.com/ovachiever/droid-tings --skill scvi-tools-ovachiever

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill provides scvi-tools frameworks for probabilistic single-cell analysis, batch correction, and complex multi-omics modeling with scVI, SCANVI, TOTALVI, and more.

Core Features & Use Cases

Probabilistic Modeling: Variational autoencoders for scRNA-seq and multi-omics data.
Batch Correction & Integration: Robust integration across batches and studies.
Differential Expression: Bayesian testing with uncertainty quantification.
Multimodal Capabilities: TOTALVI, MULTIVI, MRVI for multi-omics data.
Advanced Workflows: Transfer learning (scArches), trajectory inference (Decipher), and marker-based cell typing (CELLASSIGN).

Quick Start

Register your AnnData object, pick a model (SCVI/SCANVI), train, and then extract latent representations for clustering and DE analysis.

Frequently Asked Questions about scvi-tools

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I correct batch effects across multiple scRNA-seq datasets?▼

Batch correction in scRNA-seq uses probabilistic models like scVI to integrate data across studies and experimental conditions. Register your AnnData object with batch covariates, train the model, and extract corrected latent representations for downstream analysis.

Can I integrate multimodal single-cell data from CITE-seq and scATAC-seq together?▼

Yes. scvi-tools provides TOTALVI and MULTIVI for multimodal integration across protein, RNA, and chromatin modalities. These models handle modality-aware likelihoods and generate unified latent representations for joint analysis.

What's the best way to perform differential expression testing on single-cell data?▼

Bayesian differential expression testing in scvi-tools provides uncertainty quantification on latent representations. Train a probabilistic model, then extract posterior samples to compute credible intervals and effect sizes without multiple-testing corrections.

Do I need GPU acceleration to run scvi-tools models?▼

GPU acceleration is required for efficient training of variational autoencoders on large single-cell datasets. The setup-train-extract API leverages GPU compute; CPU execution is feasible for small datasets but significantly slower.

How do I annotate cell types using marker genes with single-cell data?▼

CELLASSIGN within scvi-tools enables marker-based cell type annotation by learning a probabilistic mapping between genes and cell types. Combined with scVI latent representations, it handles uncertainty in assignment across batches and conditions.

Can scvi-tools handle spatial transcriptomics data alongside standard scRNA-seq?▼

Yes. scvi-tools supports spatial transcriptomics analysis through its multimodal framework, enabling integration of spatial location information with gene expression for joint tissue mapping and cell typing across modalities.